The conversation delves into the limitations of static benchmarks in evaluating large language models, arguing that dynamic game environments like Diplomacy offer richer, more realistic assessments of AI behavior—especially in negotiation, strategy, and social reasoning. Alex explains how games expose subtle differences between models and enable fine-tuning for specific behaviors, turning prompt engineering into a form of artistic translation. Inspired by child psychology and theories of play, he advocates for game-based training to foster continual learning in AI—something current systems lack. Humor-driven games may also reveal new frontiers in nuance and creativity. Beyond improving AI, these interactive experiences make the technology more accessible and less intimidating to users. Looking ahead, Alex sees transformative potential in software, life sciences, and education, while emphasizing that young people entering the workforce should embrace AI as a mentor rather than fear replacement.